research article performance evaluation of moving small...

Research ArticlePerformance Evaluation of Moving Small-Cell Network withProactive Cache

Young Min Kwon,1 Syed Tariq Shah,1 JaeSheung Shin,2

Ae-Soon Park,2 and Min Young Chung1

1College of Information and Communication Engineering, Sungkyunkwan University, 2066 Seobu-Ro, Jangan-Gu,Suwon, Gyeonggi-Do 16419, Republic of Korea2Mobile Access Research Division, Electronics and Telecommunications Research Institute, 138 Gajeongno, Yuseong-gu,Daejeon 34129, Republic of Korea

Correspondence should be addressed to Min Young Chung; [email protected]

Received 21 January 2016; Revised 20 June 2016; Accepted 21 June 2016

Academic Editor: Juan C. Cano

Copyright © 2016 Young Min Kwon et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

Due to rapid growth in mobile traffic, mobile network operators (MNOs) are considering the deployment of moving small-cells(mSCs). mSC is a user-centric network which provides voice and data services during mobility. mSC can receive and forward datatraffic via wireless backhaul and sidehaul links. In addition, due to the predictive nature of users demand, mSCs can proactivelycache the predicted contents in off-peak-traffic periods. Due to these characteristics, MNOs consider mSCs as a cost-efficientsolution to not only enhance the system capacity but also provide guaranteed quality of service (QoS) requirements to moving userequipment (UE) in peak-traffic periods. In this paper, we conduct extensive system level simulations to analyze the performanceof mSCs with varying cache size and content popularity and their effect on wireless backhaul load. The performance evaluationconfirms that the QoS of moving small-cell UE (mSUE) notably improves by using mSCs together with proactive caching. We alsoshow that the effective use of proactive cache significantly reduces the wireless backhaul load and increases the overall networkcapacity.

1. Introduction

Due to the increasing number of smart phone devices anddata services, the users demand formobile data traffichas alsoincreased. Global mobile traffic will nearly increase tenfolduntil 2018 [1]. To accommodate this emerging demand ofdata traffic, mobile network operators (MNOs) have alreadyadopted advanced communication techniques such asorthogonal frequency division multiple access (OFDMA),multiple input multiple output (MIMO), and carrier aggre-gation (CA). It is possible to make the spectrum efficiencyreach its theoretical limit in 4G mobile network by usingthese technologies. However, the networks only imple-menting these advanced radio access and transmissiontechnologieswill not be able to accommodate the tremendousincrement of mobile traffic and it may exhaust the availablesystem capacity of 4G mobile networks. Thus, MNOs haveconsidered heterogeneous networks (HetNets) in order to

continuously improve the systems capacity by adding morebase stations [2, 3].

The HetNet terminology indicates that various types offixed small-cells (fSCs) such as pico- and femtocell coexist ina macrocell. fSCs can share the traffic overload of macrocellby providing mobile services to densely populated areas suchas hotspots [4]. However, fSCs using wired backhaul havedrawback in terms of signaling overheard, infrastructure cost,and mobility [5]. When many fSCs densely exist in cellularnetworks, frequent handovers occur between macrocell andfSCs [6]. For successful handover, both base stations ofmacrocell and fSCs should exchange control messages viawired backhaul comprised of several network entities [7].Thus, dense deployment of fSCs increases signaling load inthe wired backhaul. Secondly, existing fSCs require wiredbackhauls such as optical fiber or coaxial cable, in orderto connect them to the core network. Laying these wired

Hindawi Publishing CorporationMobile Information SystemsVolume 2016, Article ID 6013158, 11 pageshttp://dx.doi.org/10.1155/2016/6013158

2 Mobile Information Systems

backhaul is not a very cost-effective solution for MNOs.Moreover, fSCs using a wired backhaul cannot consistentlyprovide wireless broadband services to users that ride publictransportation vehicles [8]. Recently, working group (WG)of 3GPP standardization has investigated the moving cellutilizing the wireless backhaul as a solution to overcome thelimitations of fSCs [9].

In this paper, we introduce the concept of moving small-cell (mSC) with various transmission paths, that is, wirelessbackhaul, sidehaul, and caching transmission.mSCs are user-centric networks that autonomously establish connectionsbetween users and provide the voice and data services whilemoving [10]. mSCs communicate with their respective MBSsvia wireless backhaul links. mSCs can also exchange datathrough wireless sidehaul links among neighboring mSCs.Due to predictable nature of users, the nodes in the networktrack can learn and construct the users’ demand profilesin order to predict their future requests effectively. Thus,in the proposed mSC network, each mSC has a storagecapability to cache the predicted contents. The proposedcaching mechanism is proactive in principle and it aimsto anticipate users demands. It can reduce the backhaulload by saving the scarce frequency resources. Due to theseunique characteristics, mSC has several advantages overother fSCs. By supporting group handover, mSCs can reduceboth signaling overhead and handover failure probability[11]. Since wireless backhaul and sidehaul links do notrequire any additional deployment cost, mSCs can become acost-efficient solution to enhance the systems capacity [12].Furthermore, the proposed proactive caching mechanismused in mSCs can not only reduce the traffic load of wirelessbackhaul link but also guarantee quality of service (QoS)performance in peak-traffic hours [13, 14].

The deployment of mSCs can enhance system capacityand accommodate the increasing mobile traffic with reason-able cost. Instead of deploying new fSCs,mSCs can be utilizedas a cost-effective solution to solve the temporary hotspotissues. Although mSCs have many advantages in terms oftraffic distribution and system capacity, their performance islimited due to cotier interference among neighboring mSCs.Since mSCs accommodate all the data traffic of wirelessbackhaul link, wireless sidehaul link, and proactive contentcache, it is obvious that the performance of mSC is affectedby ratio between data traffic delivered via these various links.Thus, we have developed and conducted extensive systemlevel simulations to analyze the effect of mSCs with proactivecaching enabled in a multitier HetNet environment.

Contributions. System level simulation is one of the mostuseful methodologies to analyze the performance of vari-ous network scenarios [15]. A preliminary version of thispaper appears in the 8th ACM International Conference onUbiquitous Information Management and Communication(IMCOM), 2014 [16]. In this study, we first highlight thechallenges associated with mSCs deployment in multitierHetNets scenarios. Then, in order to exploit the advantagesof mSCs and proactive caching, we evaluate and compare theperformance ofmSCs in differentmultitier HetNet scenarios.We show the relation between contents popularity, cache size,

and operating modes and their positive effects on overallnetwork performance.

The rest of the paper is organized as follows. Section 2presents the previous studies related to mSCs and proac-tive caching. In Section 3, we introduce the proposed mSCnetwork, its architecture, and proactive caching mechanismused. Section 4 contains the detailed performance evaluationof proposedmSC network and Section 5 provides the conclu-sion of this paper.

2. Related Works

Due to unprecedented growth in mobile data traffic, networkdensification and modification in its current architectureare inevitable. In order to maximize the reuse of availablefrequency spectrum, introducing HetNets is one of the keysolutions. HetNets can accommodate the growing demandof data traffic by deploying more small-cells in a given area[2, 17, 18]. In [19], Dhillon and others have proposed atractable model for a K-tiers downlink HetNet. It showsthat in an ideal HetNet scenario, beside severe interference,the network densification can still significantly enhance theoverall network capacity. In order to provide better andreliable network services to moving users, the use of mSCshas been proposed, studied, and evaluated in [20–24].

The authors in [20] have shown that, in a coveragelimited scenario the use of coordinated and cooperative relaysin public vehicles can significantly improve the networkexperience of on-board moving users. In [8, 21–23], Suiand others have studied performance of moving relay node(MRN), which is a type of mSCs, in cellular networks. MRNsare deployed in public transportation vehicles such as trains,trams, and buses in order to provide wireless broadbandservices to moving UE. Since MRN uses wireless backhaullink to connect to MBS, it can reduce the cost of wiredbackhaul link. In addition, by supporting group handover ofall on-board UE, MRN can significantly reduce the signalingoverhead and probability of handover failure. Compared toMBS, MRN is very close to its UE; therefore it can enhancethe signal quality of the respectiveUE in access link.However,the performance of MRN mainly depends on the capacity ofwireless backhaul link [21, 22]. Since the capacity of wirelessbackhaul link is normally limited, it is difficult to increase theoverall network capacity by deploying large number ofMRNssignificantly.

The ability to predict user demands and recent develop-ments in context awareness and data storage has enabled thefuture networks to proactively cache the popular contentsin advance [25–28]. The proactive caching technique insmall-cells will not only reduce the backhaul load but alsoguarantee the QoS requirements in peak-traffic periods.In [25], Tadrous and others have studied the concept ofproactive resource allocation by utilizing the predictabilityof user behavior for load balancing. Authors in [26] haveproposed the idea of femtocaching in fSCs with very limitedbackhaul bandwidth and large storage capacity. Authors in[27] have studied the asymptotic scaling laws of caching inD2D communications. In their proposed distributed cachingscheme, users store the popular contents and forward them to

Mobile Information Systems 3

MBS

mSC

MUE

Wireless sidehaul link (f1)

Access link for mSUE (f2)

mSC

mSUEEdge cache

Edge cache

mSUE

Access linkfor MUE (f1)Wireless

backhaul link

(f1)

Figure 1: Moving small-cell network.

other users by usingD2Dcommunications. Bastug andothersin [28] have examined two cases of proactive caching. First,in order to reduce the backhaul load, they have proposeda mechanism, which proactively caches the popular files inoff-peak hours (e.g., at night) proactively. In second case,based on the social structure of the network, the proposedscheme predicts the set of potential users who can proactivelycache and distribute the popular contents utilizing D2Dcommunications.

Nonetheless, these studies on proactive caching have onlyconsidered the fSCs (pico- and femtocells)which usually havewired backhauls and do not have any backhaul bandwidthconstraint.Moreover, they also rarely consider themobility ofeither small-cells (picocells and femtocells) or users (D2D).These key aspects are the motivation behind this paper andthe aim of this is to study the role of proactive caching inmSCs.

3. Proposed Moving Small-Cell Network withProactive Cache

3.1. Network Architecture. The proposed mSC network con-sists of four network entities, MBS, macrocell UE (MUE),mSC, and moving small-cell UE (mSUE) as shown inFigure 1. MBS in mSC network provides wireless access linkand backhaul link connections to its servingMUE andmSCs,respectively. EachmSC is amoving small-cell, which provideswireless broadband services to its serving mSUE in accesslinks. To communicate between mSCs directly, mSCs canalso establish wireless sidehaul connections with their neigh-boring mSCs. Based on measurement information, the MBSis also responsible for radio resource management of bothwireless backhaul and sidehaul links of mSCs. Furthermore,in our proposed mSC network, each mSC has the abilityto cache popular contents. If mSUE requests contents thatare already stored in the cache of its connected mSC, themSC directly sends the contents to its mSUE. More detail onproactive caching is given in the next section.

As discussed earlier, due to wireless backhaul and side-haul connectivity, mPCs can be deployed on moving vehiclesto provide enhanced network services to moving UE. Itis obvious that, instead of deploying large number of fSC,

mSCs are the cost-efficient technique to serve moving UEand increase the overall network capacity. In order to avoidsevere interference between MUE and mSUE, both MBSsand mSCs in the proposed scheme use different frequencybands of 2.0GHz and 3.5GHz in their access links, respec-tively. Figure 2 shows the proposed channels and frequenciesassignment scheme for wireless backhaul/sidehaul and accesslinks of mSCs, MUE, and mSUE, respectively. In mSCnetwork, in-band full duplex transmission may be used forwireless backhaul link. Thus, for wireless backhaul transmis-sions, mSCs share the same radio resources of uplink anddownlink in 2GHz frequency band withMUE. Furthermore,mSCs also perform in-band half-duplex transmission forwireless sidehaul links, where they reuse the uplink radioresources of mSC backhaul and MUE in 2GHz frequencyband. Unlike MUE, mSUE is very close to the serving mSCs;thus the transmit power of mSC is relatively lower thanMBS.

3.2. Proposed Proactive Caching Scheme for mSC Network.It is mentioned earlier in this paper that preloading andproactive caching can significantly reduce the traffic loadon wireless backhaul link and conserve the scarce radioresources.The key issues of proactive caching are methods todecide caching data and an efficient mechanism to transmitthe selected data (preloading) [29]. This paper focuses onthe second key issue of cache preloading. We assume that,based on collaborative filtering (CF) tools [30], the MBScan effectively decide the popularity of the contents suchas video contents (e.g., TV series and advertisements), webcontents (e.g., daily news, blogs, and digests), and softwareupdate files (e.g., software drivers and patches) [31]. Thesecontents are usually time-insensitive and available long beforetheir scheduled publishing time. The effect on time-sensitivecontents has not been evaluated in this paper; it is becausewe assume that MBSs transmit the selected contents to theirrespective mSCs in off-peak period (e.g., night time). Inother words, the cache of mSCs in our proposed schemeis only updated in low traffic hours when the traffic loadon backhaul link is very low [28]. In order to continuouslyupdate the cache with time-sensitive popular contents, a full-time dedicated backhaul link is required. However, due toscarce availability of the radio resources, it is not feasibleto fully dedicate certain backhaul resources only for cachemanagement.

In order to make the preloading scheme more efficient,the MBS transmits the popular contents to mSCs in twopossible modes: broadcasting andmulticasting. If the contentfiles are equally popular among all mSCs in the network,the MBS will broadcast the selected contents to all mSCs inthe network. Similarly, if different content files are popularamong different mSCs, the MBS will make groups of mSCswith same interest and it will multicast the desired contents toeach particular group. Furthermore, inmulticast modemSCsof one group can exchange their cache contents with nearestneighboring mSC of other groups via sidehaul link. In otherwords, if the requested contents are available in neighboringmSCs, theMBS will provide the necessary information (mSCID, radio resources for sidehaul, and so on) of that particularmSC in order to establish sidehaul link. In our proposed


Frequency

Radio resources for DL Radio resources for UL

mSC

mSUEmSUE

Access link for mSUE

(DL)

mSUEmSUE

Access link for mSUE

(UL)

MBS

mSC

Wireless backhaul link (DL)

MUE mSC MUE

Access link for MUE

(DL)

Frequency

Radio resources for DL Radio resources for UL

mSC

mSC

FrequencyWireless

sidehaul link

Wireless backhaul link (UL)

Access link for MUE

(UL)

2.0GHz

· · ·· · ·

3.5GHz

Figure 2: Channel assignment in mSC network.

network, mSCs can only establish sidehaul connection withneighboring mSCs that are located in the radius of 200meters. Since, sidehaul links reuse the uplink frequencies ofboth mSCs andMUE, they can significantly reduce the back-haul traffic load. Note that the aim of our proposed scheme isto evaluate the performance of a fully loaded MSC networkwith active sidehaul links and proactive cache, under theconstraint of limited wireless backhaul capacity. Therefore,in our proposed network model, we have considered thatthe number of mSCs in each macrocell and the number ofmSUE pieces in each mSC are uniform and fixed. The aimof such network model is to find the upper bound of networkcapacity. Consequently, due to these considerations the trafficconditions of an mSC in our proposed network do not varyover time and the resource allocation is static. Figure 3 shows

the proposed preloading scheme, where, during off-peakperiod, the backhaul bandwidth is divided into two parts,one for reactive backhaul traffic and the second for proactivebroadcast/multicast caching traffic.

In our proposedmSC network, the network performancedepends on three different factors: content popularity distri-bution, cache size of mSC, and the number of multicastinggroups. In this paper, popularity distributions are obtainedfrom ZipF (𝛼) distribution [32]. It has been shown in [33,34] that the global content popularity usually follows theZipF distribution. It is also shown in [34] that a simplemodel for an independent request stream following a ZipFdistribution is sufficient to capture certain asymptotic prop-erties observed at proactive caches (such as web proxies).Another reason for using ZipF distribution is its simplicity;


Table 1: Simulation parameters.

Parameter Macrocell Moving small-cellCell layout Hexagonal grid, 3 sectors per site Uniform random distributionRadius of cell 166m (ISD = 3𝑅 = 500m) 10mNumber of cells 7 sites 20∼100Access link

Carrier freq. 2GHz 3.5GHzBandwidth 10MHz 10MHz

Tx power 46 dBm (downlink) 23 dBm (downlink)23 dBm (uplink) 23 dBm (uplink)

Wireless BH/SH linkCarrier freq. — 2GHzBandwidth — 10MHz

Tx power — 46 dBm (downlink of BH)23 dBm (uplink of BH, SH)

Antenna pattern Three-sector (2D) Three-sector (2D) (BH)Omnidirectional (2D) (SH)

Antenna height MBS: 25m mSC: 2mMUE: 1.5m mSUE: 1.5m

Mobility model MUE: random walk model mSC: random walk modelmSUE: group moving

Sidehaul connection —Contents and distance based

connection(max distance: 200m)

Number of UE pieces per cell 300 − (2 ⋅ 𝑀) 2 per mSC𝑀 = the number of mSCs

mSC mSUE

MBSmSC mSUE

(3) D2D traffic between mPCs

(2) Proactive caching traffic

(1) Wireless backhaul traffic

Broadcast/multicast traffic

Preloading scheme for mSC

Figure 3: Off-peak time proactive caching scheme for mSC net-work.

we believe that the complexity cost of other machine learningalgorithms will overburden the MSC network which havelimited computational capabilities. In ZipF distribution, 𝛼 isthe characterization exponent that ranges from zero to one.Moreover, it is obvious that the performance of mSC networkis decidedly dependent on cache size (𝑆). Huge cache sizecan significantly reduce the backhaul load and improve theQoS of mSC network. Furthermore, unlike broadcast mode,orthogonal radio resources are required for each multicastgroup transmission.Thus, the number ofmulticasting groupscan significantly affect the performance of overall network.

4. Performance Evaluation

4.1. Simulation Environment. In order to evaluate the perfor-mance of our proposed mSC network with proactive cache,we conducted system level simulations. We consider a seven-macrocell network, where each cell consists of three hexago-nal sectors. MBSs are located in the center of each macrocelland the intercell distance is 500 meters. MUE and mSCs arerandomly deployed and then they move within macrocells.Similarly,mSUEpieces are randomly anduniformly deployedand move within the coverage area of their serving mSCs. Inorder to capture the real time mobility pattern of mSCs, wehave used random walk mobility model [35]. According toour considered random walk model the moving cell (whichcan be a public transportation vehicle) travels in a randomdirection with random velocity and flight time.More detailedsimulation parameters are given in Table 1.

In our system level simulator, we have adopted ITU UMaand WINNER path loss models for macrocells and mSCs,respectively. ITU UMa model considers urban macrocellenvironment [36, 37]. Pathloss equation of ITU UMa modelis as follows:

PL = 22.0log10(𝑑) + 28.0 + 20log

10(𝑓𝑐) ,

10m < 𝑑 < 𝑑BP,

PL = 40.0log10(𝑑) + 7.8 − 18.0log

10(ℎBS)


− 18.0log10(ℎUT) + 2.0log10 (𝑓𝑐) ,

𝑑BP < 𝑑 < 5000m,(1)

where 𝑑 is distance between transmitter and receiver. 𝑓𝑐is

carrier frequency with range of 2 to 6GHz. ℎBS and ℎUT areantenna heights of BS andUE, respectively, where𝑑BP is breakpoint distance defined as

𝑑BP =4ℎBSℎUT𝑓𝑐𝑐

, 𝑐 = 3.0 ⋅ 108m/s. (2)

WINNER model provides pathloss model for small-cellswhich has low power and small coverage area [38, 39] and itspathloss equations are

PL𝐵1 total (𝑑) = max (PLfree (𝑑) ,PL𝐵1 (𝑑)) ,

PLfree (𝑑) = 20log10 (𝑑) + 46.4 + 20log10 (𝑓𝑐

2.0

) ,

PL𝐵1(𝑑) = (44.9 − 6.55log

10(ℎBS)) log10 (𝑑)

+ 5.83log10(ℎBS) + 18.38

+ 23log10(𝑓𝑐) ,

(3)

where PLfree(𝑑) and PL𝐵1(𝑑) mean free space pathloss and

pathloss for small-cell, respectively.In this paper, we have used the overall network capacity

(𝐶Total) as a performance metric, which is total sum ofmacrocell capacity (𝐶Macro) and mSC capacity (𝐶mSC) indownlink. The capacity of each cell depends on the spectralefficiency and bandwidth assigned to UE. Spectral efficiencyof UE can be obtained as the relationship between the signalto interference and noise ratio (SINR) and modulation andcoding scheme (MCS) table [36].

Let 𝑈 and 𝑀 denote the numbers of MUE pieces andmSCs deployed in each macrocell, respectively. 𝑈

𝑘denotes

the number of mSUE pieces in the coverage of mSC 𝑘. Thetotal available bandwidths in 2GHz and 3.5GHz frequencybands are 𝑊

2GHz and 𝑊3.5GHz, respectively. We define the

macrocell capacity (𝐶Macro) as the sum of all MUE capacities.Thus, it can be calculated as

𝐶Macro = (1 − 𝜌)𝑊2GHz𝑈 +𝑀

𝑈

∑

𝑖=1

MCSDL (SINR𝑖) , (4)

where 𝑖 means index of MUE attached to the MBS. 𝜌 (0 ≤𝜌 ≤ 1) depicts the ratio of radio resources for broad-casting/multicasting to overall radio resources for 2GHzdownlink.

Likewise, the capacity of mSC 𝑘 (𝐶mSC,𝑘) is also definedas the sum of all connected mSUE’s capacities. However, thecapacity of mSC depends on its transmission mode, that is,relay mode, cache mode, and mSC-to-mSC (sidehaul) mode.If mSUE requests a content file not cached in its respective orneighboring mSCs, the mSC performs relay transmission. Inthis case, themSUE receives its data viawireless backhaul linkand access link formSUE.Thus, capacity ofmSC 𝑘 (𝐶mSC,𝑘) inrelaymode is defined as theminimumvalue between capacity

Net

wor

k ca

paci

ty (b

ps)

1.2G

1.0G

800.0M

600.0M

400.0M

200.0M

0.0

No cacheZipF distribution parameter (𝛼 = 0.2)

ZipF distribution parameter (𝛼 = 1.0)

Number of mSCs per macrocell0 20 40 60 80 100

Figure 4: Network capacity varying numbers of mSCs operating inbroadcast mode.

of wireless backhaul link (𝐶BH,𝑘) and capacity of access linkfor mSUE (𝐶access,𝑘) and it can be expressed as

𝐶mSC,𝑘 = min (𝐶BH,𝑘, 𝐶access,𝑘) ,

𝐶BH,𝑘 = (1 − 𝜌)𝑊2GHz𝑈 +𝑀

MCSDL (SINRBH,𝑘) ,

𝐶access,𝑘 =𝑊3.5GHz𝑈𝑘

𝑈𝑘

∑

𝑗=1

MCSDL (SINR𝑗) .

(5)

Similarly, if mSUE requests a content file that is availablein the cache to its serving mSC, the mSC performs cachetransmission. In cache transmissionmode, themSUEdirectlyreceives its requested data from its serving mSC via accesslink. Thus, the capacity of mSC 𝑘 operating in cache modecan be determined by the capacity of its access link for mSUE(𝐶access,𝑘). On the other hand, mSCs operate in sidehaultransmission mode, if the contents requested by mSUE arenot available in its serving mSC but are available in the cacheof a neighboring mSC. The neighboring mSC delivers suchdata to serving mSC via wireless sidehaul link. The servingmSC forwards the received data to its respective mSUE viaaccess link. In this case, capacity of mSC 𝑘 (𝐶mSC,𝑘) is decidedas the minimum value between capacity of wireless sidehaullink (𝐶SH,𝑘) and access link (𝐶access,𝑘), and it can be expressedas𝐶mSC,𝑘 = min (𝛿 ⋅ 𝐶SH,𝑘, 𝐶access,𝑘) ,

𝛿 =

{

{

{

0, if sidehaul link does not exist

1, if sidehaul link exists,

𝐶SH,𝑘 = 𝑊2GHzMCSUL (SINRSH,𝑘) .

(6)

4.2. Simulation Results. Figure 4 shows the overall networkcapacity with varying number of mSCs operating in broad-cast mode. It depicts that the overall network capacity is


Net

wor

k ca

paci

ty (b

ps)

1.2G

1.0G

800.0M

600.0M

400.0M

200.0M

0.0

ZipF distribution parameter (𝛼)0.2 0.4 0.6 0.8 1.0

M (number of mPCs)20

100

(a) Network capacity

Back

haul

load

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0



100

(b) Backhaul load

Satis

fied

requ

ests

1.0

0.8

0.6

0.4

0.2

0.0



100

(c) Number of satisfied requests

Figure 5: Effect of ZipF distribution (𝛼) on overall network performance operating at broadcasting and multicasting mode.

highly dependent onnumber ofmSCs in the cell. It also showsthat the mSC with cache scenario outperforms the no-cachescenario, because most of the contents requested by mSUEare already available in the cache of mSCs. Furthermore,as the popularity of files increases (𝛼 increases.) the overallnetwork capacity also increases. It is because more mSUEpieces request the already cached files.

Similarly, Figure 5 depicts the effect of ZipF distribu-tion (𝛼) on overall network capacity, backhaul load, andnumber of satisfied requests. Two different mSC deploymentscenarios (sparse and dense) are considered. Figure 5(a)shows that, beside the inter-mSC interference, the overallnetwork capacity in dense deployment scenario (100mSCsper macrocell) is significantly higher than sparse deployment

scenario (20mSCs per macrocell). The reason is that eachmSC uses the same 2GHz frequency band in access link.In dense deployment, more mSCs reuse the same frequencyband in their access links. Likewise, Figure 5(b) depicts thatthe backhaul load significantly reduces as the file popularityincreases. Furthermore, it also shows that, in both deploy-ment scenarios, the file popularity has no major effect onbackhaul load. In this work, we define backhaul load asthe ratio of number of mSCs using backhaul link over totalnumber ofmSCs. Similarly, Figure 5(c) illustrates the relationbetween satisfied requests and file popularity. It shows that inboth deployment scenarios the number of satisfied requestsincreases as the popularity of file increases. Here the termof satisfied requests means the ratio between numbers of


Net

wor

k ca

paci

ty (b

ps)

600M

500M

400M

300M

200M

100M

0

Cache size (S)4 8 12 16

ZipF distribution parameter (𝛼 = 0.2)ZipF distribution parameter (𝛼 = 1.0)

(a) Network capacity



Back

haul

load

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

(b) Backhaul load



1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

Satis

fied

requ

ests

(c) Number of satisfied requests

Figure 6: Effect of cache size (𝑆) on overall network performance with varying ZipF distribution parameters.

satisfied requests over total number of requests. If a usersuccessfully receives a file size of 1MB within 1 second afterhis request, we call this request as satisfied one.

The effect of cache size (𝑆) on overall network capacity,backhaul load, and number of satisfied requests is shownin Figure 6. It is shown in Figure 6(a) that, with fixednumber of mSCs (in this case 20), the overall networkcapacity significantly increases as the cache size increases.Since large cache size can proactively store popular contents,they can also significantly reduce the backhaul traffic load(Figure 6(b)) and increase the numbers of satisfied userrequests (Figure 6(c)) in mSC network.

Figure 7 shows the effect of multicast groups on overallnetwork capacity. It can be observed from Figures 7(a) and7(b) that, for two different zip distribution parameters (𝛼 =0.2 and 𝛼 = 1), the broadcast mode outperforms themulticast mode. It is because theMBS inmulticast mode uses

orthogonal channels to transmit different contents to differ-ent mSC groups (in this case 2 groups), and thus it consumesmore backhaul bandwidth than broadcast mode. Figure 7(c)depicts the comparison of different resource utilization ofmSCs operating at broadcast and multicast modes. It can beobserved that in both broadcast andmulticast mode the ZipFdistribution factor plays a vital role and the backhaul loadreduces to 61% and 59% when it approaches to 1, respectively.Furthermore, the utilization of sidehaul link in multicastmode increases up to 14% when 𝛼 approaches to 1.

5. Conclusion

In this paper, we discuss the role of mSCs in future HetNetsand proposed a novel proactive caching based mSC network.We show that, by using the predictive nature of user demands,next generation networks can effectively preload their cache

Mobile Information Systems 9N

etw

ork

capa

city

(bps

)

250.0M

200.0M

150.0M

100.0M

50.0M

0.0

Number of mPCs per macrocell0 20 40 60 80 100

ZipF distribution parameter (𝛼 = 0.2)

BroadcastMulticast(a) Network capacity when 𝛼 = 0.2

Net

wor

k ca

paci

ty (b

ps)

Number of mPCs per macrocell0 20 40 60 80 100

BroadcastMulticast

ZipF distribution parameter (𝛼 = 1.0)1.2G

1.0G

800.0M

600.0M

400.0M

200.0M

0.0

(b) Network capacity when 𝛼 = 1.0

Ratio

of e

ach

tran

smiss

ion

1.0

0.8

0.6

0.4

0.2

0.0Broadcast Broadcast(𝛼 = 0.2) (𝛼 = 1.0) (𝛼 = 0.2) (𝛼 = 1.0)

Multicast Multicast

Backhaul transmissionCaching transmissionD2D transmission(c) Ratio of different transmission modes

Figure 7: Influence of multicast groups on overall network performance with varying values of 𝛼.

with popular contents and reduce the traffic data demandin peak hours. Our extensive system level simulation resultsshow that the proposed mSC network can significantlyimprove the QoS performance and overall system capacityof the network. We also show that the overall network per-formance is highly dependent on number of mSCs deployed,cache size, and content popularity. For future studies, weare aiming at incorporating the transmitted power controlschemes in our simulator, which will effectively mitigatecross- and cotier interference in mSC networks. Anotherinteresting line of investigation is to study various resourcepartitioning and scheduling schemes, which can staticallyor dynamically divide radio resources between macrocelland mSCs and reduce the interference and improve overallperformance of the network.

Competing Interests

The authors declare that they have no competing interests.

Acknowledgments

This work was supported by Institute for Information &communications Technology Promotion (IITP) grant fundedby the Korean government (MSIP) (no. R0101-15-244; Devel-opment of 5G Mobile Communication Technologies forHyperconnected Smart Services).

References

[1] Cisco, “Global mobile data traffic forecast update, 2013–2018,”White Paper, 2014.


[2] A. Damnjanovic, J. Montojo, Y. Wei et al., “A survey on 3GPPheterogeneous networks,” IEEE Wireless Communications, vol.18, no. 3, pp. 10–21, 2011.

[3] S.-P. Yeh, S. Talwar, G. Wu, N. Himayat, and K. Johnsson,“Capacity and coverage enhancement in heterogeneous net-works,” IEEEWireless Communications, vol. 18, no. 3, pp. 32–38,2011.

[4] D. Lopez-Perez, A. Valcarce, G. De La Roche, and J. Zhang,“OFDMA femtocells: a roadmap on interference avoidance,”IEEE Communications Magazine, vol. 47, no. 9, pp. 41–48, 2009.

[5] X. Ge,H. Cheng,M.Guizani, and T.Han, “5Gwireless backhaulnetworks: challenges and research advances,” IEEE Network,vol. 28, no. 6, pp. 6–11, 2014.

[6] T. Yamamoto and S. Konishi, “Impact of small cell deploymentson mobility performance in LTE-advanced systems,” in Pro-ceedings of the IEEE 24th International Symposium on Personal,Indoor and Mobile Radio Communications (PIMRC Workshops’13), pp. 189–193, London, UK, September 2013.

[7] 3GPP, “Evolved Universal Terrestrial Radio Access (E-UTRA)and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Overall description; Stage 2,” 3rd Generation Part-nership Project (3GPP), TS 36.300, v12.3.0, March 2011, http://www.3gpp.org/ftp/Specs/html-info/36300.htm.

[8] Y. Sui, J. Vihriala, A. Papadogiannis, M. Sternad, W. Yang, andT. Svensson, “Moving cells: a promising solution to boost per-formance for vehicular users,” IEEECommunicationsMagazine,vol. 51, no. 6, pp. 62–68, 2013.

[9] L. Chen, Y. Huang, F. Xie et al., “Mobile relay in LTE-advancedsystems,” IEEE Communications Magazine, vol. 51, no. 11, pp.144–151, 2013.

[10] 3GPP, “Evolved Universal Terrestrial Radio Access (E-UTRA);Study on mobile relay,” 3rd Generation Partnership Project(3GPP), TR 36.836, v2.0.2, June 2014, http://www.3gpp.org/dynareport/36836.htm.

[11] Y. Sui, Z. Ren, W. Sun, T. Svensson, and P. Fertl, “Performancestudy of fixed and moving relays for vehicular users with multi-cell handover under co-channel interference,” in Proceedings ofthe 2nd IEEE International Conference on Connected Vehiclesand Expo (ICCVE ’13), pp. 514–520, IEEE, Las Vegas, Nev, USA,December 2013.

[12] A. Enrique, “User-centric wireless local loop use-cases: a briefoverview on assumptions and requirements,”White Paper, 2011.

[13] A. Ghosh, N. Mangalvedhe, R. Ratasuk et al., “Heterogeneouscellular networks: from theory to practice,” IEEE Communica-tions Magazine, vol. 50, no. 6, pp. 54–64, 2012.

[14] E. Bastug, J.-L. Guenego, and M. Debbah, “Proactive small cellnetworks,” in Proceedings of the 20th International Conferenceon Telecommunications (ICT ’13), pp. 1–5, Casablanca, Morocco,May 2013.

[15] L. Chen, W. Chen, B. Wang, X. Zhang, H. Chen, and D. Yang,“System-level simulationmethodology and platform formobilecellular systems,” IEEE Communications Magazine, vol. 49, no.7, pp. 148–155, 2011.

[16] Y. M. Kwon, J. Shin, J. S. Kim, S.-M. Oh, M. Y. Chung, and A.-S. Park, “Development of system level simulator for evaluatingperformance of moving personalcell network,” in Proceedings ofthe 8th ACM International Conference on Ubiquitous Informa-tionManagement andCommunication (ICUIMC ’14), p. 19, SiemReap, Cambodia, January 2014.

[17] I. Hwang, B. Song, and S. S. Soliman, “A holistic view onhyper-dense heterogeneous and small cell networks,” IEEECommunications Magazine, vol. 51, no. 6, pp. 20–27, 2013.

[18] J. G. Andrews, S. Buzzi, W. Choi et al., “What will 5G be?” IEEEJournal on Selected Areas in Communications, vol. 32, no. 6, pp.1065–1082, 2014.

[19] H. S. Dhillon, R. K. Ganti, F. Baccelli, and J. G. Andrews, “Mod-eling and analysis of K-tier downlink heterogeneous cellularnetworks,” IEEE Journal on Selected Areas in Communications,vol. 30, no. 3, pp. 550–560, 2012.

[20] V. Van Phan, K. Horneman, L. Yu, and J. Vihriala, “Providingenhanced cellular coverage in public transportation with smartrelay systems,” in Proceedings of the IEEE Vehicular NetworkingConference (VNC ’10), pp. 301–308, Jersey City, NJ, USA,December 2010.

[21] Y. Sui, A. Papadogiannis, W. Yang, and T. Svensson, “Per-formance comparison of fixed and moving relays under co-channel interference,” in Proceedings of the IEEE GlobecomWorkshops (GCWkshps ’12), pp. 574–579, IEEE, Anaheim, Calif,USA, December 2012.

[22] Y. Sui, A. Papadogiannis, and T. Svensson, “The potential ofmoving relays—a performance analysis,” in Proceedings of theIEEE 75thVehicular Technology Conference (VTC Spring ’12), pp.1–5, IEEE, Yokohama, Japan, May 2012.

[23] Y. Sui, A. Papadogiannis,W. Yang, and T. Svensson, “The energyefficiency potential of moving and fixed relays for vehicularusers,” in Proceedings of the IEEE 78th Vehicular TechnologyConference (VTC Fall ’13), pp. 1–7, IEEE, Las Vegas, Nev, USA,September 2013.

[24] IEEE 802.16 BroadbandWireless AccessWorking Group, “IEEEC802.16j-07/087r2 mobile relay station operation,” IEEE, Tech-nical Report, 2007, http://www.ieee802.org/16/.

[25] J. Tadrous, A. Eryilmaz, and H. El Gamal, “Proactive con-tent download and user demand shaping for data networks,”IEEE/ACM Transactions on Networking, vol. 23, no. 6, pp. 1917–1930, 2013.

[26] N. Golrezaei, K. Shanmugam, A. G. Dimakis, A. F. Molisch,and G. Caire, “Femtocaching: wireless video content deliverythrough distributed caching helpers,” in Proceedings of the IEEE(INFOCOM ’12), pp. 1107–1115, IEEE, Orlando, Fla, USA, 2012.

[27] M. Ji, G. Caire, and A. F. Molisch, “Fundamental limits ofdistributed caching inD2Dwireless networks,” in Proceedings ofthe IEEE InformationTheoryWorkshop (ITW ’13), pp. 1–5, IEEE,Sevilla, Spain, September 2013.

[28] E. Bastug, M. Bennis, and M. Debbah, “Living on the edge:the role of proactive caching in 5G wireless networks,” IEEECommunications Magazine, vol. 52, no. 8, pp. 82–89, 2014.

[29] A. Bar-Noy, R. E. Ladner, and T. Tamir, “Optimal delay formedia-on-demand with pre-loading and pre-buffering,” Theo-retical Computer Science, vol. 399, no. 1-2, pp. 3–11, 2008.

[30] J. Lee, M. Sun, and G. Lebanon, “A comparative study of collab-orative filtering algorithms,” https://arxiv.org/abs/1205.3193.

[31] R. Vale andH. Viswanathan, “Further Efficiencies with eMBMSPreloading,” https://techzine.alcatel-lucent.com/further-effi-ciencies-embms-preloading.

[32] G. K. Zipf, “Relative frequency as a determinant of phoneticchange,”Harvard Studies in Classical Philology, vol. 40, pp. 1–95,1929.

[33] K. Shanmugam, N. Golrezaei, A. G. Dimakis, A. F. Molisch,andG.Caire, “FemtoCaching: wireless content delivery throughdistributed caching helpers,” IEEE Transactions on InformationTheory, vol. 59, no. 12, pp. 8402–8413, 2013.

[34] L. Breslau, P. Cao, L. Fan, G. Phillips, and S. Shenker, “Webcaching and Zipf-like distributions: evidence and implications,”


in Proceedings of the IEEE 18th Annual Joint Conference of theIEEEComputer andCommunications Societies (INFOCOM ’99),vol. 1, pp. 126–134, New York, NY, USA, March 1999.

[35] D. B. Johnson and D. A. Maltz, “Dynamic source routing inad hoc wireless networks,” in Mobile Computing, T. Imielinskiand H. Korth, Eds., chapter 5, pp. 153–181, Kluwer Academic,Dordrecht, Netherlands, 1996.

[36] 3GPP, “Further advancements for E-UTRA physical layeraspects (Release 9),” 3rdGeneration Partnership Project (3GPP)TR 36.814, v9.0.0, 2010.

[37] 3GPP, “Evolved Universal Terrestrial Radio Access (E-UTRA);Radio Frequency (RF) system scenarios (Release 11),” 3rdGeneration Partnership Project (3GPP), TR 36.942, v11.0.0,September 2012.

[38] IST-4-027756 WINNER II D1.1.2V1.2, WINNER II ChannelModels, 2007.

[39] D5.3: WINNER+ Final Channel Models, V1.0, 2010.

Submit your manuscripts athttp://www.hindawi.com

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttp://www.hindawi.com

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation http://www.hindawi.com Volume 2014


Applied Computational Intelligence and Soft Computing

Advances in

Artificial Intelligence


Advances inSoftware EngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

http://www.hindawi.com Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in


research article performance evaluation of moving small...

Documents